Replication-coupled repair of crotonaldehyde/acetaldehyde-induced guanine-guanine interstrand cross-links and their mutagenicity^†

Synthesis of interstrand cross-linked 15-mer duplex

The syntheses, purification and characterization of 15 mer-15 mer duplexes containing a crotonaldehyde/acetaldehyde-induced stereoisomeric guanine (N₂)-guanine (N₂) ICL, 3-(2-deoxyribos-1-yl)-5,6,7,8-(N²-deoxyguanosyl)-6(R or S)-methyl-pyrimido[1,2-α]purine-10(3H)one, or their stable 1,3-bis(2′-deoxyguanos-N²-yl)butane derivatives have been reported (18) and the same oligonucleotides were used in this study. The sequence of the duplex oligonucleotide is shown in Figure 1B. Two duplexes contained the authentic R or S stereoisomeric ICL, and a third contained the chemically stable derivative of the R isomer (Scheme 1). These ICLs were named ICL-R, ICL-S and ICL-Rd, respectively.

Construction of plasmid

The parental plasmid, pBTE, has been described (19). This plasmid replicates and is stably maintained extrachromosomally in human cells. It confers blasticidin S resistance (bla^R) on human and E. coli cells. The Blp I-PspOM I fragment (67 mer/68 mer duplex) of this plasmid was replaced with another Blp I-PspOM I fragment (94 mer/95 mer duplex) containing unique Bsa I, Afl II and BsmB I sites (Figure 1). This plasmid was named pBTEX3, and the Bsa I and BsmB I sites were used to incorporate the modified oligonucleotides. Three derivatives of pBTEX3 were also constructed: the first derivative, named pTRC, is defective in the transcription of the bla^R gene in human cells. It was constructed by removing the Avr II-Avr II fragment [390 base pairs (bps)] containing the SV40 early promoter (Figure 1A). The second derivative, named pREP, is defective in the replication in human cells. It was constructed by removing the Nsi I-Nsi I fragment (2835 bp) containing the SV40 and BK virus replication origins and the BK T antigen gene (Figure 1A). The third derivative was constructed to serve as an internal control to quantify the relative ICL repair efficiency. This plasmid was constructed by replacing the Xho I-Sac II fragment (2927 bp) containing the bla^R- and ampicillin-resistant (amp^R) genes with a kanamycin-resistant (km^R) gene cassette (960 bp) prepared by PCR amplification of the cassette in pCR4Blunt-TOPO (Invitrogen, Carlsbad, CA). This plasmid was named pINTC.

Insertion of a modified duplex oligonucleotide into plasmid

The strategy for incorporating a cross-linked oligonucleotide is shown in Figure 1B. pBTEX3, pTRC and pREP were digested with Afl II and BsmB I, and a large fragment was purified by Qiaquick PCR purification Kit (Qiagen, Valencia, CA). 5′-Phosphorylated duplex oligonucleotide was ligated to the digested vector at the BsmB I-cleaved site at 4 °C overnight (Step 1) and then digested with Bsa I (Step 2). Following the purification of a large fragment, DNA was self-ligated to form closed circular DNA (Step 3). The ligation mixture was redigested with Afl II to remove the residual parental plasmid. The modified construct was purified by ultracentrifugation in a CsCl-ethidium bromide solution (Figure 2A). The amount of DNA was quantified by a UV spectrophotometer. Purified constructs were named pICL-R, pICL-S and pICL-Rd according to the ICL inserted. The ICL was located 18 and 19 nucleotides downstream of the stop codon (TAA in Figure 1B) of the bla^R gene: the ICL site was 5′ to the polyadenylation site and therefore within the mammalian transcription unit. All experiments were conducted in the absence of the sequence homologous to the cross-linked region.

Characterization of purified constructs

DNA constructs (100 ng) were digested with Blp I and PspOM I and then incubated with exonuclease-deficient Klenow enzyme in the presence of [α-³²P] dGTP (3000 Ci/mmol, Amersham Biosciences, Piscataway, NJ) at room temperature for 20 min (Figure 1B, Steps 4 and 5). These treatments labeled DNA fragments with two [α-³²P] dGMP (shown as αα) inserted into the PspOM I-cleaved ends, generating a labeled 33 mer/32 mer duplex connected by the ICL. A labeled 33 mer was generated from control DNA, pCont. Samples were run in denaturing 20 % polyacrylamide gel.

Introduction of modified plasmid into host cells and recovery of plasmid

NER-proficient (GM637) and defective (XPA, GM04429) cells were obtained from Coriell Institute (Camden, NJ) and cultured in Dulbecco’s modified Eagle’s medium supplemented with 10 % fetal bovine serum, penicillin (100 units/mL) and streptomycin (100 μg/mL) under 5 % CO₂ at 37 °C. This XPA cell line shows less than 2 % of normal UV-induced unscheduled DNA synthesis (Coriell Institute). Cells were plated at 1×10⁶ cells in a 25-cm² flask and maintained overnight, after which they were transfected overnight with 500 ng of a DNA construct by the FuGENE6 (Roche, Nutley, NJ) method according to the manufacturer’s instruction and then detached by treating with trypsin-EDTA. A portion (0.5 – 2 %) of cells was plated in a 10-cm dish, and blasticidin S was added to the medium at 5 μg/mL on the next day. The culture was maintained until resistant cells formed visible colonies, at which time cells were stained with a Giemsa solution. Since ICL was incorporated immediately downstream of the bla^R gene, plasmid repaired without the induction of a large deletion in the ICL region establishes the antibiotic resistance. The remaining cells were plated into a 150-cm² flask and cultured in the absence of blasticidin S, allowing the collection of all types of repair events including large deletions. When cells became nearly confluent, progeny plasmid was recovered by the method of Hirt (20) and analyzed for repair events (see below).

E. coli MM1992 (as AB1157, but uvrA6, endA7::cam, mutS201::Tn5) and MM1993 (as AB1157, but endA7::cam, mutS201::Tn5) were employed. Electroporation was used to transform bacteria with 50 ng of a construct by an E. coli Pulser (Bio-Rad, Hercules, CA). Following the incubation at 37 °C for 40 min, various volumes of a transformation mixture were plated on YT (1×) agar plates (21) containing ampicillin (100 μg/mL medium) to determine the number of transformants in the mixture. Transformants were individually picked up and analyzed for repair events.

Analysis of plasmid for repair events

In the experiments with human cells, 5 ng of pVgRXR (Invitrogen), which coded for Zeocin resistance and served as an internal control for Dpn I digestion, was added to the recovered plasmid. The mixture was treated with Dpn I (1 unit) to remove nonreplicated input DNA and then used to transform E. coli DH10BMax electrocompetent cells [F⁻, mcrA, Δ(mrr-hsdRMS-mcrBC), φ80lacZΔM15, ΔlacX74, deoR, recA1, endA1, araΔ139, Δ(ara, leu)7697, galU, galK, λ⁻, rpsL, nupG, tonA)] (Invitrogen) by an E. coli Pulser. The transformation mixture was plated on YT (1×) agar plates with ampicillin (100 μg/mL medium) or with Zeocin (25 μg/mL medium). Very large deletions spanning into the amp^R gene will not be detected in this analysis. E. coli transformants were picked up individually and subjected to oligonucleotide hybridization (22) using probes shown in Figure 3A to determine DNA sequence at the ICL site. Probe CG detected error-free repair events while probe CT detected G1→T transversions. Examples of hybridization are shown in Figure 3B and C. When a colony did not hybridize to either of them, DNA sequencing was conducted. Thus, this strategy detected all types of repair events including base substitutions, frameshifts, deletions and insertions without bias.

Construction of modified plasmid

Since crotonaldehyde/acetaldehyde-induced ICLs are reversible to monoadducts (Scheme 1), the results obtained with the authentic ICLs (ICL-R and ICL-S) may represent those for both ICLs and monoadducts. Therefore, we included an irreversible model ICL (ICL-Rd) of the R isomer and compared with the results for authentic ICLs. The gel analysis of labeled products revealed two bands (Figure 2B): one corresponded to the cross-linked 65 mer, which migrated faster than single-strand 61 mer, and the other was the large fragment of the vector, which did not migrate in the gel. The control construct yielded a 33 mer as expected. When the authentic ICLs revert to their monoadducts, the modified 33 mer will be observed. The analysis did not reveal a clear 33-mer band in any lanes of modified DNAs, indicating that the authentic ICLs were quite stable during the construction (Figure 2B). However, this does not necessarily mean that they are stable in a cellular environment. The ICL-modified 15 mer/15 mer duplex also migrated faster than a single strand 30 mer (data not shown),

Role for NER in ICL repair in E. coli

In the initial experiment, we examined the effects of NER on the repair of these ICLs. When the modified constructs were introduced into the uvrA strain (MM1992), pICL-R and pICL-S yielded reduced but substantial numbers of amp^R transformants (Table 1). In contrast, pICL-Rd failed to yield transformants. When the NER-proficient strain (MM1993) was used, all constructs yielded transformants. These results imply that portions of the two authentic ICLs reverted to their monoadducts in bacteria or during storage while the model ICL-Rd did not. This result confirms the essential role for NER in ICL repair in E. coli (23) and the stability of ICL-Rd.

Role for NER in ICL repair in human cells

In experiments with human cells, a portion of transfected cells were plated in a 10-cm plate and cultured in the presence of blasticidin S (5 μg/mL medium) until cells formed visible resistant colonies (7–10 days). Since the ICL was located immediately downstream of the bla^R gene (Figure 1A), only repair events that are not associated with large deletions give rise to resistant colonies. pICL-R and pICL-S established numerous resistant cells in both NER-proficient and defective cells as in E. coli (picture not shown). In contrast to the result in E. coli, a good number of resistant cells were obtained even in NER-defective XPA cells when pICL-Rd was used (Figure 4B). A repeated experiment showed that the numbers of resistant colonies were 422 for pCont and 193 (46 %) for pICL-Rd when 0.5 % of cells were plated. These results indicate that the human NER-independent ICL repair mechanism operates without losing a large segment of DNA. In this experiment, the repair of ICL was estimated by determining the number of bla^R colonies. However, this estimation is not accurate and the repair efficiency is overestimated since host cells likely absorb more than one molecule of the construct during transfection and the repair of one of them renders the cell resistant to the antibiotic, thus leading to the overestimation. It is technically difficult to control the number of molecules absorbed to each cell during transfection. To overcome this problem, an internal control vector (pINTC) was designed, which replicates in human cells and codes for only km^R. In this strategy, the repair efficiency is determined by the ratio of the number of amp^R colonies and km^R E. coli colonies. This approach is useful when comparing results obtained in different cell lines since a transfection efficiency can be very different from one cell line to another. A mixture (500 ng:500 ng) of pINTC and pICL-Rd was transfected into GM637 and XPA cells overnight in a 25-cm² flask, cells were replated in a 150-cm² flask and cultured in the absence of blasticidin S for 4 days before collecting progeny plasmid. Following Dpn I digestion, DH10BMax was transformed with progeny plasmid and various volumes of a transformation mixture were plated on amp (for pICL-Rd) and km (for pINTC) plates. On the next day, transformants were counted and the ratio (No. of amp^R/No. of km^R) was determined. They were 0.29 (204 amp^R/714 km^R) for GM637 and 0.06 (46 amp^R/800 km^R) for XPA. Thus, the value for GM637 is almost 5 times higher than that for XPA, implying the repair efficiency in XPA cells was 20 % of that in the GM cells. In another experiment, the value was 16 %. Thus, although NER-independent repair mechanism significantly contributes to the repair, NER still plays a major role.

Fidelity of ICL repair in E. coli

To determine a DNA sequence at the cross-linked region, we employed oligonucleotide hybridization, which could detect even one-base mismatches (Figure 3) and hence allowed us to detect any sequence changes including base substitutions and large/small deletions and insertions. To reduce the number of oligonucleotide probes, we first employed a probe which detected correct repair (probe CG). A preliminary analysis revealed that the majority of plasmid hybridized to the CG probe and the sequencing of hybridization-negative plasmid identified G1→T transversions as major targeted mutations. Therefore, to determine the fidelity of the repair, we employed the hybridization with these two probes and DNA sequencing.

The analysis of plasmid revealed that the vast majority (>97%) of plasmid had the correct sequence of 5′ CG at the site of ICL for all ICLs (Table 2). There were no significant differences in the fidelity of repair or the specificity of miscoding events among the three ICLs. The frequencies of targeted point mutations were <3% and they were almost exclusively G1→T transversions. No mutations were observed at the G2 site.

Fidelity of ICL repair in human cells

Similar to the results in E. coli, the majority (>94%) of plasmid was repaired correctly. Targeted mutation frequencies were <3 % except pICL-S in GM637, where it was a little higher (5.6 %). The major targeted mutations were G1→T transversions, and limited numbers of other types of targeted mutations such as G1→A and G1→C were also observed. The three ICLs showed similar miscoding specificity in human cells. Interestingly, targeted mutations were limited to G1 in XPA cells while a limited number of mutations (G2→A and G2→C) were also observed at G2 in GM637 cells. In addition to these targeted point mutations, limited numbers of deletions and insertions, especially the former, were also detected. These deletions were ranging from 8 to 125 nucleotides, and most of them had two - four nucleotides microhomology at the junction of deletions. There were also a few plasmid in each group, whose sequencings failed. Based on their sizes, we assume that these plasmids had large deletions, which were not determined here.

Effects of transcription and plasmid replication on ICL repair

As revealed in the bacterial experiments, certain fractions of the authentic ICLs revert to their monoadducts, and this reversion confounds the study of ICL processing in cells. The model ICL, ICL-Rd, appears to be stable in cells and its genotoxic properties also appear to be similar to those of the authentic ICLs. Therefore, further studies were conducted using pICL-Rd. In our construct, the ICL-Rd was incorporated between the bla^R gene and its polyadenylation site (Figure 1). Hence, it is possible that transcription rather than replication is important to the repair in NER-deficient XPA cells. To test this possibility, the transcription-defective derivative, pTRC, was constructed (see Materials and Methods). Unmodified pTRC (1 μg) was introduced into XPA cells and cells were selected for bla^R (5 μg/mL). When 10 % of transfected cells were plated, no resistant colonies were established unlike the parental pBTEX3 that yielded an uncountable number of resistant colonies. This proved the lack of expression of the gene in pTRC. The ICL-Rd was then inserted into this plasmid and the modified plasmid, pTRC-Rd, was introduced into XPA cells. Transfected XPA cells did not give rise to bla^R colonies as expected. However, Dpn I-resistant (replicated) progeny plasmid was recovered efficiently from the transfected cells. This suggests that the transcription event is not critical to the NER-independent repair mechanism. The amp^R colomies were analyzed for the fidelity of repair. As reported in Table 2, the repair fidelity for pTRC-Rd was the same as that observed with the transcription-competent pICL-Rd. Thus, the inactivation of transcription did not influence the ICL repair. In conclusion, the transcription does not play a significant role in the NER-independent ICL repair.

In the next experiment, the replication-defective derivative, pREP, was used to examine the role for replication. Control pREP or modified pREP-Rd (1 μg) was transfected overnight into XPA cells. All transfected cells were plated with different portions, ranging from 5 to 30% per dish, in 10-cm dishes and selected for bla^R. Replication-defective control plasmid, pREP, established resistant cells (Figure 4C), but the modified construct did not yield even a single colony (Figure 4D). The total numbers of resistant colonies were more than 4000 for pREP and <20 for pREP-Rd, based on the number of resistant colonies per plate seeded with 5 % of cells. This result is in contrast to that with the replication-competent modified construct, pICL-Rd, (Figure 4B) and suggests the importance of the replication ability for the repair events. The replication-incompetent modified plasmid may be lost before or during the incorporation into chromosomes. We cannot exclude another possibility that this is due to the sensitivity difference in the detection of repair between the two types of plasmids.

ICL repair mechanism

Here we have studied how the acetaldehyde/crotonaldehyde-induced ICLs are repaired in the absence of a homologous sequence and induce mutations in cells, using a shuttle plasmid model system. To distinguish the events caused by monoadducts and ICLs, we have included a chemically stable model ICL-Rd. In E. coli, the repair of ICL-Rd absolutely required the NER function, indicating the critical role for the NER function and at the same time the stability of ICL-Rd in E. coli. Therefore, E. coli transformants obtained with the authentic ICLs in MM1992 are likely due to the reverted monoadducts. The NER function is also important in human cells as revealed by the quantitative assay, by which the repair efficiency in XPA cells was shown to be ~20 % of that in the NER-proficient cells. This is in contrast to the results obtained by another group (16, 17): the repair of a psoralen-induced dT-dT ICL (16) and a mitomycin C-induced dG-dG ICL (17), which were inserted site-specifically into plasmid, required functional NER as in E. coli. This discrepancy is explained by the difference in the ability of the plasmids to replicate in host cells: our plasmid replicates in human cells while theirs does not. These different results also support the idea that the NER-independent repair couples with replication.

Our results are consistent with one of the current models for the initial stages of ICL repair (Figure 5), in which an NER-independent repair mechanism is active on replicating DNA at the replication fork (pathway II) while NER is active on nonreplicating DNA (pathway I). When a replication fork approaches an ICL, a DNA endonuclease such as XPF-ERCC1 incises one strand 3′ and 5′ to the ICL (13). The strand incised is probably the G2-containing leading strand template since targeted mutations were limited to the G1 site in XPA cells. This implies that TLS was conducted across G1 bearing G2. This scenario also fits the enzymatic character of XPF/ERCC1. XPF/ERCC1 mutants are very sensitive to cross-linking agents while the other NER complementation group mutants are not very sensitive (12). This heterodimer incises a strand having a 3′ unpaired tail of Y-structure DNA (13, 24). This incision specificity is consistent with our results: the replication fork progresses as shown in Figure 5 in our plasmid and the G2-containing leading strand template has a 3′ unpairing tail. The incision of this strand generates a gap opposite the modified G1, across which TLS is conducted and hence targeted mutations are limited to G1 (Figure 5). In addition, the independence of this repair mechanism from transcription implies that the Y structure of DNA is not sufficient for this repair to be initiated. In the NER-dependent pathway I, the incision appears to be made on either strand since, though a limited number, targeted mutations were also observed at the G2 site in GM637 (Table 2). However, this idea requires further research since, in the NER-proficient E. coli MM1993, all targeted mutations were limited to the G1 site. Since no homologous sequence was initially available in this model system, the repair of ICL depends on TLS as previously reported (16, 17, 25).

The ICL-Rd repair observed in XPA cells provides direct cellular evidence for NER-independent ICL repair in human cells. This mechanism can not be base excision repair initiated by a DNA glycosylase. If so, the enzyme would generate an apurinic site at G1 or G2 and this lesion has to be bypassed by TLS with a preferential insertion of dCMP opposite the apurinic site to restore the original G:C pair. However, it is generally agreed that other nucleotides, especially dATP, are preferably inserted opposite the lesion in mammalian cells (26–28). Thus, the base excision repair mechanism can not explain the repair with such a high fidelity. Furthermore, we have obtained the evidence for the involvement of ERCC1 in this repair (unpublished data).

Translesion events across authentic and model ICLs

The results of the E. coli transformation experiments indicate that the authentic ICLs revert to their monoadduct forms (Scheme 1, Table 2). Therefore, the results in MM1992 (uvrA) should represent the fidelity of TLS across the G1 monoadducts. This TLS is quite accurate with the insertion of the correct dC opposite the monoadducts, and targeted G1→T transversions were observed at a frequency of <2 %. In MM1993, the results may reflect TLS across both the G1 monoadducts and “half-excised” ICL. The results in this strain were similar to those in MM1992, suggesting that the TLS across the “half-excised” ICL adducts is also quite accurate as observed for pICL-Rd, and dA was occasionally inserted opposite these lesions.

In human cells, the results obtained in XPA cells represent the TLS across monoadducts and the replication-coupled repair intermediate for the authentic ICLs. TLS across the monoadducts has been characterized in human cells (22) and simian kidney cells (29). The study with human cells (22) reported that the S monoadduct was more miscoding than was the R monoadduct (10% vs 5%). Interestingly, pICL-S is somewhat more miscoding than is pICL-R in the two human cell lines. This may reflect the contribution of those stereoisomeric monoadducts to the miscoding events here. The TLS across the repair intermediate (pICL-Rd under XPA in Table 2) is also mostly accurate with occasional misinsertion of dA opposite the intermediate. Therefore, G1→T transversions observed with pICL-R and S in XPA cells may be induced by both types of lesions. In GM637 cells, in addition to the TLS across these lesions, TLS across NER intermediates may also occur, totaling three different TLS. As shown in Table 2, the results in GM637 are not very different from those in XPA cells. Therefore, all of these results suggest that the fidelity of TLS across these three types of lesions is similar in human cells though the TLS mechanisms may be very different from each other: correct dC is preferentially inserted opposite these lesions with occasional insertions of incorrect nucleotides, mainly dA. The DNA polymerase(s) catalyzing these TLS is not known. However, since the “half-excised” G1-G2 ICL adduct is very bulky, it is very likely that the responsible polymerase is a Y family DNA polymerase (30) and/or pol ζ. Indeed, the involvements of pol η (17) and pol ζ (31, 32) have been recently reported.